updated dependencies

mpdf is back to a release tag (my changes have been merged). minimum PHP
version is 8.2 now, just like DokuWiki.

require allowdebug for debug output requested via URL

The debug URL parameter dumped the raw mPDF HTML and bypassed the cache for any
visitor. It is now honored from the URL only when core allowdebu

require allowdebug for debug output requested via URL

The debug URL parameter dumped the raw mPDF HTML and bypassed the cache for any
visitor. It is now honored from the URL only when core allowdebug is enabled;
the config/programmatic path stays trusted.

show more ...

skip forbidden pages and distinguish them from empty selections

Collected pages the user may not read are dropped from the export and remembered
separately, so a selection where every page is forbid

skip forbidden pages and distinguish them from empty selections

Collected pages the user may not read are dropped from the export and remembered
separately, so a selection where every page is forbidden can be told apart from
one where nothing was selected and reports "no read access" instead of "nothing
selected". The forbidden message no longer lists the page ids or references the
removed 'Skip Forbidden Pages' option; its outdated translations are dropped so
they get re-translated.

show more ...

give user feedback for empty or invalid PDF exports

Collectors used to return an empty page list for a missing page, a missing
namespace or an empty selection, which then failed silently or crashed

give user feedback for empty or invalid PDF exports

Collectors used to return an empty page list for a missing page, a missing
namespace or an empty selection, which then failed silently or crashed when the
first page was accessed. They now throw a translated ExportException instead,
which the action handler turns into feedback: a flash message and redirect for
regular navigations, or an HTTP error with the message in the body for the
background download BookCreator uses. Dynamic message parts are HTML escaped
because that body is injected into the page via a jQuery html() sink.

show more ...

Fix failing tests

fix config attrbutes and extend tests

Some more tests testing rendering functionality

Again mostly LLM generated but manually checked and corrected.

pass the Config object into the renderer

This streamlines the configuration to a single point of entry. The
renderer was the last bit not adhering to that.

Overhaul chapter/header counting in renderer

Using static variables makes it impossible to automatically test this
correctly.

The PDF renderer is now a singleton. The Writer class initializes it
an

Overhaul chapter/header counting in renderer

Using static variables makes it impossible to automatically test this
correctly.

The PDF renderer is now a singleton. The Writer class initializes it
anew for each export, but during the export, the same renderer is
reused.

show more ...

introduce a first test for rendering - testing numbered headlines

The test passes, but I am not 100% sure the behaviour is correct yet.
Static variables in the renderer might be a problem. Not teste

introduce a first test for rendering - testing numbered headlines

The test passes, but I am not 100% sure the behaviour is correct yet.
Static variables in the renderer might be a problem. Not tested with
multiple documents yet.

Currently render options need to be set via global conf and are not
passed from the Config object.

show more ...

Added tests

These tests have been initially LLM written, I went through each of them
to make sure they make sense and fixed what I thought needed fixing.

However coverage is still not great and sho

Added tests

These tests have been initially LLM written, I went through each of them
to make sure they make sense and fixed what I thought needed fixing.

However coverage is still not great and should be improved in the
future.

show more ...

fixed remaining sort test

update general test

updated tests for Media resolving

require autoloader during test

Usually the autoloader is initialized fromt the DokuPDF constructor.
This is not called during individual ImageProcessor class tests.

rector and php sniffer cleanups

fix ImageProcessor test

The problem was that the test relied on the existence of a smiley image
that has recently be removed in master.

refactor the test classes

Note: the ImageProcessor test is currently broken (but was before).

Update test to include new link attribute data-wiki-id

no longer differentiate between natural and pageorder sorting

There is no reason not to always sort naturally.

mix pagenames and namespaces when sorting

As discussed in https://github.com/splitbrain/dokuwiki-plugin-dw2pdf/pull/399#issuecomment-624030852

properly sort pagenames by namespace. fixes #308

This ensures pages within the same namespace are put together, with the
start page being alwas first.

tests: fix class loading in tests

The DokuImageProcessorDecorator is now namespaced,

refactor: make getImage decorator testable and add tests

bookmarklevels may rise by 'only' +1

It seems that bookmarklevels in pdf should only increase by +1
Of course this creates a mismatch between actual levels and
bookmarklevels. This patch flattens th

bookmarklevels may rise by 'only' +1

It seems that bookmarklevels in pdf should only increase by +1
Of course this creates a mismatch between actual levels and
bookmarklevels. This patch flattens the nesting at the original level
were the mismatch started.

Fixes #218

show more ...